Instance task_state remains in 'deleting' state if Compute server is down

[Unmesh Gurjar]

Scenario: If the Compute server where the instance is running goes down, and a delete request for the instance is received, the instance state remains in 'deleting' state.
Verify the response in the database or dashboard.

Expected Response: The vm_state of instance must be 'error'.
Actual Response: The vm_state remains in the 'active' state and task_state is 'deleting'.

Branch: master

The Nova API must check if the Compute server (hosting the instance) is up, and if it is down, update the vm_state state in database to 'error'.

[Johannes Erdfelt (johannes.erdfelt)]

I don't understand this expected behavior. Why would an instance be in error if it's running fine?

[Unmesh Gurjar (unmesh-gurjar)]

I agree that the instance is running fine. However, until the Compute server comes up, the instance will remain in 'deleting' state. A user has no information about the reason the instance is in 'deleting' state for that long (from looking at the dashboard).
I have linked this bug to a blueprint for compute instance clean up service.

[Mandar Vaze (mandarvaze)]

linked to blueprint https://blueprints.launchpad.net/nova/+spec/compute-instance-cleanup-service

[Sean Dague (sdague)]

Any progress so far?

 -your friendly neighborhood bug triager

[Sumant Murke (sumant-murke)]

I tried to reproduce the scenario and found out that the current behavior of the system is a valid scenario.
When a compute server is down and a delete request is sent the instance is marked for deletion and whenever the compute server is up and running the periodic task deletes the instance.
I feel that this bug is been fixed and is no longer a valid scenario and should be marked as invalid.